The single most useful idea I have taken from more than two decades in obesity care is simple. The drug or the device on its own is never enough. What makes the difference is what I call the behavioural layer.
The behavioural layer is the structured set of human-facing support, content, interactions, and design choices that wrap around a drug or device to help patients change their behaviour and sustain that change over time. It sits alongside the clinical pathway. It is what makes the experience specific to each patient, tailored to their circumstances, their progress, and their barriers.
GLP-1 medicines are a good example. They produce substantial weight loss, but in real-world use more than half of patients discontinue within the first year. People stop early, before they see the full benefit, for a mix of reasons, many of them behavioural: difficult side effects, unrealistic expectations, waning motivation, cost, and the everyday obstacles that make lasting change hard, from emotional eating to social and environmental pressures.
This is why I treat obesity as a chronic condition that needs sustained management rather than a short intervention. Pharmacotherapy can start the change. Whether it lasts depends on adherence, activity, nutrition, and psychological support. Improving adherence, appropriate persistence, and long-term maintenance is where a lot of the remaining value in GLP-1 care now sits.
The evidence points the same way. Before the current wave of conversational AI, we took a behaviour change programme that worked in person and delivered it asynchronously through health coach-led messages in an app. We built it, evaluated it, and published the work in JMIR Formative Research. The intervention was valued by patients and associated with sustained weight loss and psychological wellbeing at twelve months, delivered entirely through messaging. The lesson was not that the app was clever. It was that a well-designed behavioural layer produces outcomes the medication alone does not.
For teams building GLP-1 products, this reframes the problem. The hard part is not prescribing or fulfilment, difficult as those are. The hard part is designing the support that keeps someone engaged and safe through onboarding, plateaus, side effects, and the moments where life gets in the way. That is behavioural work, and it is where AI, used well, can now help at a scale that human coaching alone never could.